how hard/easy is it to parallel mosfets?

Thread Starter

strantor

Joined Oct 3, 2010
6,782
I've never used a MOSFET. The first time I've had a need for them is now for trying to make a motor controller for a big truck starter motor (not modifying a vehicle, just using vehicle parts for something else), so I will need hundreds (possibly >1000A) of amps capacity. I have read that MOSFETS are more easily paralleled than IGBTs because of their positive temperature coefficient (less chance of frying them) vise the IGBT's negative temperature coefficient. I have also read that there are problems encountered when paralleling MOSFETS handling currents like this because of the extremely massive PCB traces needed and also in trying to get them to turn on/off at the same time.
I have this MOSFET in mind (10 in parallel for 1320A capacity, just to be safe).
To get around the need for acres of PCB space, I am thinking of soldering the leads directly to thick bus bar or large gauge wire. Will that cause me a problem? Capacitance?
How do I get them to turn on/off at the same time?
What kind of driver do I need? Will one driver drive all 10?

FYI this is not an H-bridge; only one direction rotation needed.
 
Lateral mosfets dont need source resistors.
Vertical mosfets do as one will take all the current and blow up.

I am fascinated you want to use mosfets when a solenoid will do the job much cheaper.
 

praondevou

Joined Jul 9, 2011
2,942
Paralleling positive temp coefficient Mosfets is no problem, I did it several times. However 10 MOSFETs take up a lot of space, so gate traces from the driver will have quite a difference in their length, that introduces inductances and can lead to non-equal switching behaviour.
However this is more important if they are PWM-controlled. Are they?

Do you want to use this MOSFET because you have it at hand or why do you want 500V breakdown voltage?

If you can live with other packages, less max voltage you could do it with less MOSFETs or IGBTs.
IGBTs exist also with pos temp coefficient, although I never parallel more than two...
 

Thread Starter

strantor

Joined Oct 3, 2010
6,782
Lateral mosfets dont need source resistors.
Vertical mosfets do as one will take all the current and blow up.

I am fascinated you want to use mosfets when a solenoid will do the job much cheaper.
Sorry, I should have mentioned that this is to a PWM controller (hopefully operating >18khz) so solenoid isn't really an option.
You've thrown me some more google topics: "vertical Vs lateral MOSFETS"; never heard of this.
Paralleling positive temp coefficient Mosfets is no problem, I did it several times. However 10 MOSFETs take up a lot of space, so gate traces from the driver will have quite a difference in their length, that introduces inductances and can lead to non-equal switching behaviour.
However this is more important if they are PWM-controlled. Are they?

Do you want to use this MOSFET because you have it at hand or why do you want 500V breakdown voltage?

If you can live with other packages, less max voltage you could do it with less MOSFETs or IGBTs.
IGBTs exist also with pos temp coefficient, although I never parallel more than two...
I am planning to get a 24V starter motor like those used in diesel trucks and double, possibly triple the voltage; so I suppose I only need a MOSFET with max of 72V. The reason I specified that MOSFET is because if the starter motor works out, I may want to expand later to something more exciting without having to rebuild my controller and might need more voltage. I could always do things the logical way and start off small though I suppose.

As far as the inductance and non-equal switching behaviour, what could I do to minimize the effect of the inductance? One thing I was thinking, since the MOSFETS may end up soldered to bus bar instead of a PCB (still not sure if that's a good idea or not...?), possibly cutting and and soldering 10 identical length insulated wires to the gates and terminating them together at one point. will it work?
 

Thread Starter

strantor

Joined Oct 3, 2010
6,782
Has anyone here done anything like this and have pictures and/or a schematic of it? I've looked all over the net and found people have done it but not eager to throw out any useful info.
 

Thread Starter

strantor

Joined Oct 3, 2010
6,782
update: found something useful. Turns out I wasn't too far off the deep end with my Bus bar idea. I think I might be thinking overkill with the 10 MOSFETs, because that's what he's using, and he's using it in a car.
 

praondevou

Joined Jul 9, 2011
2,942
in my other company the mosfets were soldered on a pcb but current was delivered to them via copper bars soldered onto the pcb tracks, it was dcdc-converter with 3kW input, so average input current at 20V was about 150A, however peak current was about 2.5x higher... still much lower than your 1000A. so it should be possible, yes.
In the link you found the guy exploded his Mosfets, you should avoid doing this...:)
There are certainly lots of controllers but I had good experiences with the uc2825A, but I'm not sure if this works for a motor... these controllers have pulse-by-pulse current limiting capability. Maybe the motor gurus here can give you better/other options.

make the gate tracks/wires as equal in length/short as possible.
Also, for these high currents a good layout of the power part becomes very important.
 

Thread Starter

strantor

Joined Oct 3, 2010
6,782
Thanks praondevou.
I was wondering about the capacitors in that guy's design. He doesn't mention them on that page, but elsewhere in another design (different capacitor, but used in the same way) he states that they are...
used to supply ripple current to smooth the power flow coming from the batteries. Panasonic capacitors were used due to the good value and reasonably good ripple current capabilities. These offer about 2 amps of ripple current each at 20kHz, 70°C.
I was thinking that power flow from batteries should be relatively stable. I would understand the capacitors if it were coming from a recitfier, but it's batteries. I don't get it. do you have any light to shed on that?
 

shortbus

Joined Sep 30, 2009
10,045
You really should get your motor and measure the current draw on it. Most starting motors show a high amp load because they are used at near stall conditions. In free running, like your vehicle at speed, the amp draw will be much less. I'm pretty sure the HP/amp rating listed for them is a stall rating, not a full RPM rating.

Couldn't you use a GTO or IGCT instead of mosfets or IGBT's? Thats what is used in mining equipment on the DC traction and other machinery.

Another thing you have to watch for is that even though mosfet/IGBT in that type package have really great looking specs in that type package, the specs mean nothing. If you do some checking what the data sheets don't/won't tell you is that the real limit to them is around 75 Watts. No matter what other volt and amp rating is shown. To get any better rating you need to go to an 'iso-top' package.
 
Last edited:

Thread Starter

strantor

Joined Oct 3, 2010
6,782
You really should get your motor and measure the current draw on it. Most starting motors show a high amp load because they are used at near stall conditions. In free running, like your vehicle at speed, the amp draw will be much less. I'm pretty sure the HP/amp rating listed for them is a stall rating, not a full RPM rating.

Couldn't you use a GTO or IGCT instead of mosfets or IGBT's? Thats what is used in mining equipment on the DC traction and other machinery.

Another thing you have to watch for is that even though mosfet/IGBT in that type package have really great looking specs in that type package, the specs mean nothing. If you do some checking what the data sheets don't/won't tell you is that the real limit to them is around 75 Watts. No matter what other volt and amp rating is shown. To get any better rating you need to go to an 'iso-top' package.
Thanks for the infor about the iso top package.
Never heard of a GTO; newark appears to only sell one, with no data sheet, and mouser doesn't appear to sell any at all. wikipedia says this:
GTO thyristors suffer from long switch off times, whereby after the forward current falls, there is a long tail time where residual current continues to flow until all remaining charge from the device is taken away. This restricts the maximum switching frequency to approx 1 kHz.It may however be noted that the turn off time of a comparable SCR is ten times that of a GTO[dubious – discuss]. Thus switching frequency of GTO is faster than an SCR.
The reason I wanted the PWM switching frquency >18khz is so that it would be inaudible. at 1000hz or below (assuming...) my motor would have a signature scream, which I guess wouldn't be a bad thing, just would be cool if it was silent (ok, quiet). Also according to wikipedia they have a 3.4V drop. if I'm running hig amps, that's going to be a high heat dissipation problem. not too big of a problem though with a big heat sink I guess. would be nice to have only one switch instead of ten. do you know where I can get one?

same for IGCT, can't find them, where are they hiding????
 

tom66

Joined May 9, 2009
2,595
At 100A each MOSFET will dissipate 390W! You will probably need liquid cooling to get rid of the 3.9kW of heat produced in those devices.

Are you using a 12V or 24V motor? A 30V MOSFET will be okay. (If 12V, look for a 20V MOSFET.) Here's one which can handle 79A. http://www.onsemi.com/PowerSolutions/product.do?id=NTD4904N

It's SMD (DPAK), but will not be difficult to solder as it is very large. It's got a 3.9mohm resistance, and will require a heatsink; dissipating ~25W at full load. However, a passive heatsink or air cooling will be all that is needed to get rid of the full ~300W. A large heatsink, of course.

At 1000A your best bet is to get your traces plated, allowing additional current flow, and to use both sides of the board, if available. I have designed a board to pass 300A briefly (5 secs max) through a trace; it was just 50 mil thick. 200 amps continuously can be done at 85 mils.
 

Thread Starter

strantor

Joined Oct 3, 2010
6,782
Dangit, leave to math to rain on a parade.

My motor will most likely be a 24V truck starter motor and I will likely run it at 48 to 72V (will have to see what I can get away with). so I guess I'm looking for a 80V MOSFET or better. Was actually hoping to go higher (300V?) initially so I can use a higher voltage motor in the future but kinda gave up on that for now. maybe I'll make a better controller later.
as far at plated traces, could I just melt a river of solder onto it? Actually I'm still not even sold on using a PCB at all. I want to investigate the feasibility of soldering MOSFET leads directly to bus bar and if that's no good, then I'll look more into beefing up PCBs.
 

Thread Starter

strantor

Joined Oct 3, 2010
6,782
Another thing you have to watch for is that even though mosfet/IGBT in that type package have really great looking specs in that type package, the specs mean nothing. If you do some checking what the data sheets don't/won't tell you is that the real limit to them is around 75 Watts. No matter what other volt and amp rating is shown. To get any better rating you need to go to an 'iso-top' package.
When you say that, are you referring to using them with no heat sink? I was planning on using a hefty heat sink an possibly a fan; would that change anything? looks like the guy in my previous link (using massive heat sink) got away with using them (well he got about 750A out of 10 paralleled before they blew). he was using a different MOSFET but same package.
 

praondevou

Joined Jul 9, 2011
2,942
5 of these would do nice don't you think?
http://www.mouser.com/ProductDetail...=sGAEpiMZZMvECErq9cesgJdUgk/FlBvJLyWwAWC6CdI=

wouldn't even need a PCB, just bolt bus bars directly to them?
What a monster :p Yes, it looks very powerful, 5 of them should be more than sufficient.

.....
I was thinking that power flow from batteries should be relatively stable. I would understand the capacitors if it were coming from a recitfier, but it's batteries. I don't get it. do you have any light to shed on that?
You create ripple current when you switch on /off your mosfet bridge with PWM. When you don't have capacitors, the switching current has to pass through the whole length of the cable coming from the battery. These cables act as inductors too which, at the moment where you turn off the PWM, can lead to high voltage spikes on the mosfets (thinking of a h-bridge here). the cables would also be a wonderful antenna, especially for such a high current. So the switching current has to come from capacitors mounted near the power mosfet bridge.
 

CraigHB

Joined Aug 12, 2011
127
as far at plated traces, could I just melt a river of solder onto it?
Solder is only about a tenth as conductive as copper so using a river of solder on your traces is not going to be highly effective in comparison. I think the RoHS solder is somewhat more conductive though. Best thing to use is a copper bus bar either soldered to a PCB or bolted down to an insulator.
 

Thread Starter

strantor

Joined Oct 3, 2010
6,782
You create ripple current when you switch on /off your mosfet bridge with PWM. When you don't have capacitors, the switching current has to pass through the whole length of the cable coming from the battery. These cables act as inductors too which, at the moment where you turn off the PWM, can lead to high voltage spikes on the mosfets (thinking of a h-bridge here). the cables would also be a wonderful antenna, especially for such a high current. So the switching current has to come from capacitors mounted near the power mosfet bridge.
from the pics, it looks like He's using 470μF Aluminum Electrolytic Capacitors. Is that the best type for the job? How do you figure he sized those?
 

Thread Starter

strantor

Joined Oct 3, 2010
6,782
SHORTBUS! HELP! I've become aroused by these GTOs but can't find any. If I find one for sale, it's a replacement part for a proprietary drive system so no data sheet available or if I find one with a data sheet, it's from the manufacturer and can't purchase online. They want you to contact them for ordering (which in my experience means that they only want to sell in bulk to a business address on corporate account, and not mail a single unit to a home address).
where on earth do you get them?
I think one alone could make super cool one-of-a-kind controller with limitless (well outside my limits) capacity! I want! I want!
 

tom66

Joined May 9, 2009
2,595
Dangit, leave to math to rain on a parade.

My motor will most likely be a 24V truck starter motor and I will likely run it at 48 to 72V (will have to see what I can get away with). so I guess I'm looking for a 80V MOSFET or better. Was actually hoping to go higher (300V?) initially so I can use a higher voltage motor in the future but kinda gave up on that for now. maybe I'll make a better controller later.
as far at plated traces, could I just melt a river of solder onto it? Actually I'm still not even sold on using a PCB at all. I want to investigate the feasibility of soldering MOSFET leads directly to bus bar and if that's no good, then I'll look more into beefing up PCBs.
You want to get the traces plated with tin or gold in the factory. This can be done by exposing the soldemask on the side of the traces; they will then be plated just like pads will be.

With a higher voltage motor, your MOSFETs need a higher Vds and their on-resistance and gate charge increases, making them less efficient at the same current. STE250NS10 are 0.0055 ohm devices, but at 220A each will dissipate ~266W, so you'll need a very good heatsink.
 
Last edited:

praondevou

Joined Jul 9, 2011
2,942
SHORTBUS! HELP! I've become aroused by these GTOs but can't find any.
I think one alone could make super cool one-of-a-kind controller with limitless (well outside my limits) capacity! I want! I want!
from Wiki: "GTO thyristors suffer from long switch off times, whereby after the forward current falls, there is a long tail time where residual current continues to flow until all remaining charge from the device is taken away. This restricts the maximum switching frequency to approx 1 kHz."

They also have quite different driver requirements.

I would stick to a Mosfet or IGBT with pos temp coefficient.

For the capacitor sizing have a look HERE. Usually the manufacturer already designes and identifies the suitable caps for switching applications, so low ESR, high ripple current, high temperature etc....
 
Top